Graded build 1 of 4 in the University of Texas at Austin – McCombs post-graduate program in AI & Machine Learning: Business Applications (September 2026), completed individually. A drivetrain-fault classifier trained on 131,760 ten-minute sensor readings across 15 turbines over a two-month window, on a course-supplied scenario and dataset. A tuned Decision Tree scored a suspiciously perfect 1.00 recall against 0.44 precision on validation; it was rejected as a likely temporal shortcut rather than treated as a win, and a tuned XGBoost model (0.87 recall / 0.61 precision on validation) shipped instead, holding at 70.5% recall / 57.1% precision on held-out test data. Recommended as a human-reviewed screening tool, not an automatic dispatch trigger: at that precision roughly four in ten flags are false alarms. The dataset is course-provided and is not republished; the full code notebook and the slide deck are. Write-up, notebook and slides: https://averyresume.com/mccombs.html